In one of my project I am using ARM cortex m3. Here I am trying to print a float variable value into a string variable ex. sprintf(sVar, "%f", 22.23); the value returned in sVar is never correct. But if I declare a float global variable, not even use this variable it starts printing value correctly into the sVar variable. Can anyone explain what might be going wrong here? THanks
using a float will pull the floating point libraries into the link. just using "%f" in a printf may not be sufficient to do so.
I understand that this might be happening in the background but I fail to understand why a local float defined does not solve this. It requires a global definition to resolve the issue. I have seen examples from Keil website for printf and that doesn't do anything different from what I've done. Can you suggest libraries that I should include manually to remove this such dependency because I tried few but that didn't solve the problem. Appreciate your help.
Note that a float is 32 bit large. So when you have an immediate float value as parameter, the compiler will most probably implement identical code as if it pushed a 32-bit integer. Nothing for the linker to see, to understand that floating point support is needed.
This is similar to the old Borland Turbo C days where "everyone" wondered why printf() wasn't able to print floating point numbers.
a local float (if it's unused) will simply get optimised away. so no need to include any support for it. a global is far less likely to get optimised out in this way; unless extreme global optimisation is carried out.
More importantly a local variable will have a random initialized value unless explicitly defined.
If you're printing something out, or calling another external routine with it as a parameter, it's not going to get optimized away